Stepan Stepanishchev created FLINK-40023:
--------------------------------------------
Summary: HiveCatalog throws TableNotPartitionedException on
partitioned non-hive table
Key: FLINK-40023
URL: https://issues.apache.org/jira/browse/FLINK-40023
Project: Flink
Issue Type: Bug
Components: Connectors / Hive
Reporter: Stepan Stepanishchev
If we try to store the non-hive partitioned table (e.g.
'connector'='filesystem') using HiveCatalog, then the table metadata is stored
inside hive metastore without information about partition keys. If we will try
to invoke operation that requires partitioned table we get
TableNotPartitionedException.
How to reproduce:
CREATE CATALOG myhive
WITH (
'type' = 'hive',
'hive-conf-dir' = '/opt/client/Hive/config',
'hive-version' = '3.1.0'
);
USE CATALOG myhive;
create table test_table(
uuid varchar(20),
name varchar(10),
dt varchar(20)
)
PARTITIONED BY (dt)
with (
'connector' = 'filesystem',
'path' = 'hdfs://hacluster/tmp/test_table',
'format'= 'csv'
);
SHOW PARTITIONS test_table;
[ERROR] Could not execute SQL statement. Reason:
org.apache.flink.table.catalog.exceptions.TableNotPartitionedException: Table
default.test_filesystem in catalog MyCatalog is not partitioned.
SHOW PARTITIONS should not throw TableNotPartitionedException for a table
created with PARTITIONED BY. If no partition instances are registered in the
catalog, returning an empty result would be acceptable. Moreover, default
catalog will also return an empty result for filesystem connector. So should
HiveCatalog.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
